draw_focal_loss_value.py

def focal_sigmoid_cross_entropy_with_logits(labels: tf.Tensor,
                                            logits: tf.Tensor,
                                            gamma: float = 2.0,
                                            alpha: float = 0.25):
  pred_sigmoid = tf.nn.sigmoid(logits)
  pt = (1 - pred_sigmoid) * labels + pred_sigmoid * (1 - labels)
  focal_weight = (alpha * labels + (1 - alpha) * (1 - labels)) * tf.math.pow(pt, gamma)
  loss = tf.nn.sigmoid_cross_entropy_with_logits(labels, logits) * focal_weight
  return loss


def dev_focal_loss_val():
  import tensorflow as tf
  import matplotlib.pyplot as plt

  k = tf.keras
  kl = tf.keras.layers
  K = tf.keras.backend

  logits = tf.range(-20, 20, 0.01)
  labels = tf.ones_like(logits)
  gamma=2
  alpha=1
  foloss= focal_sigmoid_cross_entropy_with_logits(labels,logits,gamma=gamma,alpha=alpha)
  bceloss= tf.nn.sigmoid_cross_entropy_with_logits(labels,logits)
  plt.plot(tf.nn.sigmoid(logits).numpy(),foloss.numpy(),label='focal loss')
  plt.plot(tf.nn.sigmoid(logits).numpy(),bceloss.numpy(),label='bceloss')
  plt.hlines(0.2,0,1,colors='r', linestyles='--',label='0.2')
  plt.legend()
  plt.xlabel('confidence')
  plt.ylabel('loss value')
  plt.title(f'focal loss gamma : {gamma} alpha : {alpha}')
  plt.ylim((0,5))